On Fri, Dec 25, 2020 at 09:13 PM, KWKloeber wrote:
Since this was the free service that was better than sliced bread, I think Mark might provide an explanation as to how the cost is closely related to the # of members. i.e., How does 401 members cost 55¢ more? What costs money, seems logical to me, is activity, not so much the number of members.Just thought folks might like to know a little more about how this business works. I don't know about Groups.io, but I have experience with the economics of providing online services. The costs are usually stepwise, not linear and not easily allocated to the number of customers.
Putting it more concretely, suppose you provide an online service to 500 customers that costs you, the service operator, $1000 per month and you charge your customers $4 per month, giving you $1000 net per month with 500 customers. (Revenue=500x$4= $2000. Cost=$1000. Net = Revenue - Cost = $2000 - $1000 = $1000).
Providing services to 501 customers is also $1000 per month. In fact, providing services to 1000 customers is likely to also be very close to $1000 per month. Why? Because you use the same infrastructure (software, compute, storage, network) to provide for 500 as 1000 customers. There is little difference between using a sliver of available capacity and using most of available capacity. That means that adding customers costs you, the service operator, almost nothing. For every dollar your customers pay, that dollar goes straight to your net. At 1000 customers, your net is now $3000. (Revenue = 1000 x $4 = $4000. Cost = $1000. Net = $4000 - $1000.) Not bad.
But when the 1001th customer arrives your capacity is exceeded, performance tumbles and you start having outages. Then you have a problem. You have to add capacity or lose customers. Enough capacity so that you don't have to run through the same drill next week, but if your revenue is based on the number of subscribers, that 1001th subscriber will only add $4 to your net, which won't buy any additional capacity at all. So you dip into your reserves (you do have reserves don't you?) and raise your capacity by a factor of ten. The economy of scale makes your cost for 10,000 customer capacity only $5000, not the $10,000 you might expect. That looks great because at 10,000 customers you would net $35,000 per month (Revenue = 10,000 x $4 = $40,000. Cost = $5000. Net = $40,000 - $5000 = $35,000.) Quite a jump from $3000 a month net.
Holy Mackerel Batman! This looks like a money machine!
The catch at 1001 customers, you are losing money to the tune of $996 per month. (Revenue = 1001 x $4 = $4004. Cost = $5000. Net = $4004 - $5000 =. $996 LOSS). Your net is underwater until you take on an additional 200 customers. (Rev = 1200 x $4 = $5000. Cost - $5000. Net =$5000 - $5000 = 0). There's money to be made, but you have to get your paying customers above 1200 before your losses drive you out of business. A traditional business solution in this kind of bind is the "loss leader" to hustle customers in the door. In fact, free services are more business savvy than traditional loss leaders because there's less affect on cash flow. And you can sometimes use your free customers as a sort of buffer for testing while shielding your paying customers. Adding 10 free customers does not increase your cost, while attracting 20 paying customers with 10 free gets you closer to that magic point where your net begins to emerge from the sea of red.
Real life is more complicated than my cooked example and cloud computing has taken some of the sting out of excess capacity (even cloud fees are usually stepwise), but the general principle is valid: online costs tend to stay constant as the number of users increase, then jump disproportionately when a threshold is crossed. Business models must eventually reflect that reality. Most past online service business models, including Mark's, have been less than future-proof and times are changing. There are lots of ways this nut can be cracked, and I expect to see a lot of different ways in the next year or so.
The discussion here is all toward better models, and that is good. Best, Marv